Other Parts Discussed in Thread: TPA3245, TPS565201, PMP30092,
I'm designing a SMPS for an audio application, but I'm struggling to make it matching the spec.
On the schematic/pcb there are provisions for other features like SSR, overvoltage/overtemp protection, which are not used at this stage.
Universal input 85-264Vac.
Main output (+VPW) is 30V 5A nominal power, and it's used for a TPA3245 class-d amp. When unit is in stand-by, no power is drawn from this rail, however it cannot exceed 32V to avoid damage to the audio amp.
+VD is an intermediate out from which is derived a 5V for logic through a DC-DC converter. 5V must be able to provide at least 1A when the main output is unloaded and >3A when unit is ON.
+VA/-VA are auxilliary outputs for classic audio opmaps. Post-regulation is done on the mainboard. +VA/-VA should be withing the range 15V-25V. Max loading: 200mA when unit is ON, disconnected when unit is off.
No load power must be <0.25W at high mains.
The main issue of the prototype is very poor cross-rail regulation.
(Vin=230Vac) |
NO LOAD on +VPW |
2.7A LOAD on +VPW |
Note |
+VPW (main out) |
30V |
30.6V |
|
+VA |
18V |
>32V |
10Kohm load |
-VA |
-18V |
>-32V |
10Kohm load |
+VD |
8.7V |
18V (clamped by 15V TVS) |
15Kohm load - near TPS565201 max Vin |
+VAUX |
11.5V |
15.5V (clamped by 15V zener parallel to C27) |
|
Power input |
230mW |
98.5W |
|
As you can see, secondary outputs can even double the nominal voltage when main output is fully loaded, with the risk of failure for those components not rated that such high voltages.
I wasn't expecting very tight regulation on secondary rails, but in this case that's too poor.
On top of that, if I load just the +VD rail, the main output tends to rise dangerously as well, which is not acceptable.
One of the things I'm going to change is to derive the 5V directly from +VPW main out, which I will test soon (using a different buck IC).
How to improve cross-rail regulation? I know that this is mostly related to transformer construction, so I'm attaching spec of the one used in the prototype.
Any suggestion on winding sequence, use of internal shielding, turns spacing etc is very welcome.
Would it be preferable to switch to secondary side regulation in this case? Obv. using a different PWM controller.
Also, the RCD/TVS snubber gets very hot when fully loading the main output.
I've tried to play with RCD values, but couldn't find a good compromise with mosfet Vds peak and no-load input power yet. Any suggestion how to proceed to tune values?
When a diode/TVS solution is preferable over classic RCD snubber?
I had a look at PMP30092, but I couldn't understand why the transformer is 250uH but if you type same design spec in the UCC2863x excel calculator, it suggests only 100uH for the transformer.
Is it generally preferable to work with higher primary inductance if TX construction allows for it? (more turns, different core, etc)
To conclude, as my suggestion for TI, I would like to see the option for multi-output topology in webench.